Portugal

What the record shows

Portugal holds this atlas's purest case of a city that was only ever a rendering, and a capital whose recurring weakness is the opposite of that failure — a very good layer built above things that wobble. The two sit an hour apart and are worth reading together.

PlanIT Valley was to be a from-scratch city near Porto for roughly 225,000 people, run by an "Urban Operating System" with billions of sensors. The money needed for full build-out never materialised and nothing was built. It is the smartest city never built, and its lesson is the one that generalises furthest in this atlas: a city is not a software product, and without residents, financing and political anchoring even the most complete urban operating system stays a rendering. Set it beside Masdar City and Egypt's new capital and a pattern emerges about what greenfield announcements actually deliver on the ground.

Lisbon did the opposite and built the integration layer for a city that already existed. PGIL has pulled real-time data from around forty municipal and partner systems into one operations dashboard since 2017 — traffic, bike-share and parking use, waste-bin fill levels, pollution and noise — and the same data feeds a university-built mobility-planning tool and an EU-co-financed data lab. It is a genuinely working platform. The interesting problem is what sits underneath it.

GIRA shows the shape of it. The municipal parking company's docked bike-share opened with ten stations in 2017 and now spans all twenty-four parishes; every bike carries GPS, the app shows live per-station availability and e-bike battery level, and a command centre uses that feed to dispatch rebalancing vans. It has logged more than 12.27 million trips and 46.4 million kilometres. And the operator's own 2022 budget cut planned investment in the system by about 75 per cent, so by May 2025 independent reporting found barely a third of the claimed fleet of around 2,000 bikes actually on the street. Real-time availability data means very little when the fleet itself is unreliable — the dashboard is accurate and the service is not. The drainage programme is the same gap in concrete: the first tunnel's boring finished in July 2025, a second is now due in 2029, and a separate pilot of ten flood and flow sensors has been running since 2023. A master plan on paper is not drainage in the ground, and even a well-funded flagship adaptation project slips years past its own deadlines.

Two entries turn on who is allowed to reuse what, and Lisbon answers the question twice in opposite directions. The municipal bus and tram operator has published a static schedule feed since 2020 and tracks its fleet by GPS through a platform upgraded in 2022, with real-time positions reaching Google Maps in December 2023 and its own app in June 2025 — then restricted outside developers' access, calling its GPS feed "public, not free", after an independent tracking app used it. The metro operator next door runs an open API that outside developers build on. Publishing a schedule is not the same as opening real-time data, and the difference here is policy rather than capability — the same split this atlas found between Istanbul's bus operator, which publishes its feeds, and Ankara's, which does not. The city's own catalogue goes the other way: Lisboa Aberta has published municipal datasets since February 2016, past 360 by 2023, each one tagged as a public-domain dedication or a standard attribution licence — a clean position, and a reminder that a city can be open in one office and closed in another.

The record's strongest instrument is also its most easily overlooked. Lisbon has run 80 multi-sensor stations — 658 sensors in total — since March 2021, tracking air pollutants, noise, weather and hourly traffic, contracted to named vendors and part-funded by an EU programme, with readings published openly. It has something almost nothing else in this atlas has: an independent benchmark, a master's thesis at a Lisbon university that checked nine months of hourly readings from the low-cost sensors against five reference stations of the official network. And its entry carries a warning worth generalising — because the network is one of roughly forty systems feeding the PGIL dashboard, a purpose-built deployment with its own contractors and its own funding is easy to mistake for a feature of the platform that displays it.

The last entry is the one that should trouble anyone who builds participation tools. Lisbon was the first European capital to run a citywide participatory budget, from 2008, taking proposals online and by text message; twelve editions through 2021 drew nearly 7,000 citizen proposals and funded 162 projects. No edition has run since. The platform did not break and the residents did not lose interest — a change of administration withdrew the political backing, and a 2025 attempt to restart it did not produce a thirteenth round. The portal is still online, and nothing new gets proposed or voted on it. If the entries below agree on one lesson, it is that in Portugal the software has rarely been the fragile part: what fails is the fleet under the dashboard, the tunnel behind the plan, and the mandate behind the platform.